With major investments pledged by the likes of Amazon Web Services, data centres will become even bigger business in Australia. However, they present both risks and opportunities.
Data centres have evolved rapidly since they originated in the 1940s to house massive mainframe computers, although their primary purpose has not changed. Used for powering and cooling servers, these dedicated spaces now have a hand in almost every aspect of a company’s critical operations, including its shared email servers, e-commerce platforms, remote working software, supply chains, financial transactions, and AI and machine learning. They also act as storage vaults for an organisation’s confidential information.
Beyond the rise of the internet and cloud computing, it is AI — and GenAI specifically — that is fuelling the enormous demand for data centres. In June, Amazon Web Services (AWS) announced it will invest $20b in data-centre infrastructure in Sydney and Melbourne by 2029. It represents the largest technology investment of its kind in Australian history and could help position the nation as a global AI and technology leader, as well as meeting the soaring domestic demand for cloud computing and AI.
It is critical for a board to treat their organisation’s data centre arrangements with priority and an ongoing focus, says Rik Irons-Mclean, chief technology officer (CTO) of Enterprise Commercial at Microsoft Australia and New Zealand.
“As Australian businesses accelerate their use of AI and cloud solutions, directors need to recognise that data centres are no longer just technical infrastructure — they’re strategic assets that hold the potential to accelerate an organisation’s ESG impact. And from a governance perspective, secure, compliant infrastructure and transparent ESG reporting are now board-level priorities. Directors who engage with these dimensions will be better positioned to lead responsibly and deliver long-term value.”
New opportunity
In the past, data centres primarily supported cloud computing, enterprise software, data storage and web hosting. These services were optimised for general-purpose computing. With the rise of AI, data centre investment is increasingly oriented around two distinct computational phases — training and inference.
The training phase involves building an AI model by learning patterns from large datasets. It is resource-intensive and requires prolonged processing time. “For example, training the initial version of the AI model behind ChatGPT consumed about 552 MWh of electricity — roughly the annual usage of 120 US households — plus hundreds of thousands of litres of water for cooling,” explains Professor Liming Zhu, research director at Data61, the data and digital specialist arm of the CSIRO.
After training comes the inference phase. This includes answering questions, generating content or images and scanning documents for summaries and insights. Its energy consumption is typically even higher than the training phase — and this is where most opportunities for innovation are currently available.
“As AI becomes more widely integrated across applications, data centre usage for the inference phase is expected to outpace that of the training phase, eventually becoming the dominant source of data centre demand,” says Zhu.
He adds that the shift has design implications. Inference and training require different chip architectures and system optimisations, so investment strategies may need to reflect this divergence. “While Australia may have missed the early race to train large language models (LLMs) and GenAI, it is well-positioned to lead in trust, security and system-level innovation at inference time, with specialised data centre investment.”
According to CBRE’s Australia’s Data Centres 2024 report, the investable universe of the Australian data centre sector is forecast to nearly double from $23b to $40b in the next four years. Australia is currently ranked fifth in the world for its data centre built-out capacity (megawatts) behind the US, China, UK and Germany.
“Sydney and Melbourne are the main hosts to most of the large data centres in Australia, mainly due to infrastructure and business activity. However, the current pipeline of committed capacity indicates significant investment being made in Perth, Brisbane and Canberra,” said CBRE’s head of data centre research, Sass Jalili, in a media release.
Environmental impacts
Data centres consume up to 50 times more energy than a typical commercial building, according to the US Department of Energy, meaning significant emissions relative to other industries, plus indirect emissions generated by the manufacturing and transportation of equipment.
Data centres under construction in Australia will have operational loads of 100MW-plus and proposed projects could see that figure soar to one gigawatt-plus (1GW).
The US Energy Information Administration says 1GW (one billion watts) could power nearly 900,000 households for a year.
“The high electricity demand requires that data centre operators source renewable energy to reduce their emissions,” says Alex Moffatt, head of APAC Data Centre Capital Markets and Advisory at commercial real estate company Cushman & Wakefield. “Equally, investment in new equipment to produce more efficient cooling will reduce the electricity required and also reduce emissions.”
There is a growing focus from policymakers and stakeholders on how data centres consume energy. Earlier this year, the federal government mandated new power usage targets for fresh data centre facilities. This measure will save millions of kilowatt-hours over time. Less energy used means less stress on local power grids.
Centre operators have also contracted PPAs (power purchase agreements), which enable the data centre to source renewable electricity. They and their partners are also investing directly in battery energy storage systems (BESS). Both of these measures significantly reduce emissions.
“Operators aim to maximise their reliance on renewable energy sources to reduce requirements for onsite generation, which is a last resort,” says Moffatt. “This focus on renewable energy mitigates environmental impact, provides long-term cost stability and enhances brand reputation.”
Similarly, water use effectiveness (WUE) is becoming a critical consideration for data centres in Australia and beyond. WUE measures water consumption in litres relative to IT energy used in kilowatt hours. A 100MW facility can consume up to two million litres of water per day.
Social component
With their high energy consumption, data centres can create tensions with local communities, including pressure on utility infrastructure and upward pressure on electricity prices. “Long-term benefits must be weighed against these disruptions,” says Zhu.
Moffatt notes initiatives like Microsoft’s training programs in partnership with TAFE NSW and its Datacentre Academy support equal digital access and broader STEM education. “Data centres are considered leaders in Australia for social initiatives,” he says.
Sustainability-linked bonds are another prime example of social impact. AirTrunk is a hyperscale data centre specialist focused on the Asia-Pacific and Japan region. AirTrunk is one of the largest issuers of sustainable financing in the global data centre industry with a total financing platform of more than $18b.
The recent acquisition of AirTrunk for $24b by Blackstone Inc has made the asset management giant the world’s largest data centre provider.
“Data centre operators also invest significantly in new equipment to improve efficiency and sustainability of their facilities,” says Moffatt.
He adds that when evaluating data centre operators, directors would be advised to consider location, cybersecurity, tier rating, system redundancy, electricity costs, physical security, technical support and certifications. “These factors directly impact business continuity and reputation.”
The idea of a net positive impact (NPI) is gaining currency. This extends the idea of sustainability to create an overall benefit for the environment and society.
“Organisations and projects adopting this approach seek to go beyond minimising harm and strive to make a measurable and positive contribution to the wellbeing of communities and ecosystems,” says Moffatt.
Microsoft is also on the case. “Environmentally, we’re designing facilities to be carbon-negative, water-positive and zero-waste by 2030, with innovations like zero-water cooling and circular centres that recycle over 90 per cent of server components,” says Irons-Mclean.
Comparisons matter
Zhu notes that the environmental impacts must be put in perspective. For example, asking an AI to summarise 100 documents or generate a technical report can save hours of human labour at a computer.
“This could offset the energy otherwise used in traditional workflows, such as by manually reading and writing on high-powered devices,” he says. “Ultimately, the value of the work enabled by AI and comparison to existing workflow and other typical human activities should factor into the environmental assessment.”
Emissions may be reduced in time by centralisation. At the moment, many generative large AI models such as ChatGPT are general-purpose. A single trained model can support thousands of use cases across organisations, eliminating the need for millions of separate smaller AI models that would each have incurred their own training cost.
“This centralisation arguably could reduce overall emissions,” says Zhu. “By focusing on data centres, directors can help their organisations leverage AI effectively while mitigating associated risks and challenges.”
Moffatt adds boards play a crucial role in driving the integration of ESG principles into data centre construction, expansion and operations.
“This often includes establishing an ESG committee, given that up to 70 per cent of the ongoing operating costs of the data centre are related to electricity and water consumption.”
Security and governance
Custodians of masses of sensitive information, data centres are a prime cybercrime target. They are a “national security business” under the Security of Critical Infrastructure Act 2018, which requires disclosure of ownership and reporting of cyber incidents. It permits the federal government to prevent or impose conditions on foreign investment in data centres on national security grounds.
“Requirements for physical and digital security at data centres are higher than most other social infrastructure,” says Moffatt. “Regular reviews of security infrastructure, employee training and incident response plans are critical. Upholding strong data privacy and security practices is considered fundamental to responsible data centre operation.”
This article first appeared as 'Founts of knowledge' in the November 2025 Issue of Company Director Magazine.
Latest news
Already a member?
Login to view this content